Research On Security In Vehicular Ad Hoc Networks

With the rapid development of national economic and urbanization process, increasing vehicle population has brought intense pressures and a series of contradictions. It leads to many problems in current road transport system, such as traffic congestion, accidents, air pollution, energy consumption, etc. Vehicular ad-hoc networks(VANET) consist of on board units, road side units and back-end servicers. Various kinds of applications, such as collision avoidance, incident alert, and traffic condition perception, can be implemented by providing drivers with beyond-visual-range traffic information. Because VANET is applied in road traffic system, VANET attacks not only could destroy functionality of VANET applications, but also cause traffic disorder, even incidents in extreme cases. Thus, it is very important to propose reliable VANET security mechanisms before its large-scale deployment. This dissertation focuses on VANET security problems, and gives detailed classify of VANET attacks and analysis of its characteristics. Also, the difficulties and the challenges to solve these problems are given. It explores a series of key problems, such as secure data aggregation, false messages detection, and obtains some theoretical and technological achievements.Considering security threats in VANET data aggregation, this dissertation proposes spatial-temporal correlation based secure data aggregation approach. Although VANET data aggregation can save communication resource, it leads to serious security threats: the signatures of data provider on atomic reports are abandoned in aggregation procedure in which non-repudiation is damaged. It leaves a backdoor for attackers to inject false reports. Most of existing detection schemes need a strong assumption “honest majority”, thus they fail to defense effectively colluded attacks launched by multi adversaries. Since traffic variables, such vehicular speed, density, exhibit generally strong spatial correlation and temporal autocorrelation, this dissertation proposes a spatial-temporal correlation based secure data aggregation mechanism. In the proposed approach, the location that a data anomaly is found is viewed “invisible location”, the value of traffic variables in the location are estimated by an artificial neural network based estimator using collected data from adjacent spatially and temporally road segments. Their reliability is evaluated by the deviation degree between reported and estimated value, and highly deviated data aredetected as false data. For reducing the bandwidth consumption of detection algorithm, the dissertation models the attacker-detector rational interactive decision process as a two-player incomplete information dynamic game, analyzes mix-strategy perfect Bayesian equilibrium, and proposes optimum detector’s detection strategy with maximum individual payoff. The simulation results show that the proposed approach can provide better performance on detection accuracy and robustness against collision attacks, and achieve trade-off between accuracy and communication resource consumption.Focusing on false message attack problem launched inside adversaries, the dissertation proposes secure multi-source data fusion based false alert message detection approach. Observe vehicles fuse data collected from multi-kinds of onboard sensor using belief quantify and combination approach provided by Dempster-Shafer evidence theory, recognizes traffic patterns, and evaluates the reliability of received alert messages by calculating basic belief assignment(BBA). For enlarging observation range and enhancing detection accuracy, BBAs generated by multi observers are fused. Because colluded attackers can inject forged BBAs to support false alert messages, evidence distance concept is used to quantify the similarity of two BBAs and an outlier detection algorithm is employed to detect the false BBAs. The performance of proposed approach is evaluated by the simulation experiment. The results show that the proposed approach can provide enhanced robustness against collusion attacks and better detection accuracy.Because selfish nodes can leads to negative effect on multi-nodes cooperative security mechanisms, the dissertation proposes a coalition game based VANET node credibility evaluation approach. Because various kinds of security mechanism, such as misbehavior detection, secure routing protocol, need multi-vehicles cooperative with each other, selfish nodes that don’t cooperate can lead to decreased system performance. We assume that vehicle nodes are rational individual with being able to intellectually make decision to maximize its benefit. The cooperation behaviors among vehicle nodes are modeled using cooperation game theory, and a coalition game based node credibility establishment approach is proposed in the dissertation. Multi nodes form a group in ad-hoc style, each group member measure received radio signal strength of other members and a cross-check is used to detect the false position of attackers and false data injected by them. A payoff allocation function satisfying “core” allocation condition is proposed to stimulate nodes in joining coalitions. The dissertation proves the proposed function satisfies individualrational condition and efficiency condition, the optimum strategy of nodes is to cooperate with each other, the formed coalition has stability. The simulation results show that the cooperation behavior of nodes can reduce the impact of random error of signal error strength on accuracy of position verify, and the detection rate of false position surpass 90%, and robustness against collusion attacks is enhanced.For reducing the risk of privacy disclosure in V2 I communication, the dissertation proposes a network coding based privacy protection V2 I communication scheme. By traffic analysis attack, eavesdropping adversaries can recovery packet forwarding paths, track vehicular traces and compromise multi-pseudonym based privacy protection mechanism in V2 I communication sessions. The dissertation proposes a network coding based privacy protection V2 I communication scheme that exploits slicing, coding and buffering mechanism, eliminates correlation of packers on size, content and forwarding time, and hide effectively forwarding path of packets. Hence, it can protect vehicular identity and location privacy. The security analysis shows that the proposed scheme can provide communication confidentiality, anonymity, unlinkability and untracebility. The simulation results show that the proposed scheme can reduce significantly the computational overhead comparing to existing multi encryption based secure routing protocol.